CVE-2026-5757 Details

A vulnerability allowing unauthenticated remote information disclosure has been identified in Ollama's model quantization engine. This issue arises from an out-of-bounds heap read/write vulnerability, which allows an attacker to read and exfiltrate the server's heap memory. The vulnerability is present in Ollama, an open-source tool for running large language models locally on various operating systems, including macOS, Windows, and Linux. The issue is caused by the quantization engine's lack of proper bounds checking on tensor metadata from user-supplied GPT-Generated Unified Format (GGUF) files. Exploitation of this vulnerability could lead to unauthorized access to sensitive data, further system compromise, and stealthy persistence.

As of now, no patch is available for this vulnerability. However, access to the model upload functionality should be restricted or disabled, especially in environments exposed to untrusted users or networks. If model uploads are necessary, only accept models from trusted sources and apply appropriate validation controls.